Optimization of robotic shaping control for non-rigid material with safety constraints

Robotic milling has emerged as a prominent research area for complex shaping owing to the inherent flexibility of manipulators. The investigation of robotic milling for rigid material has widely reported. However,because of the inherent attribute of rigid material, these methods mainly concentrate on the position and pose control of manipulators during task processing, lacking of reaction to material state, such as temperature and elastic deformation. This oversight becomes problematic when dealing with nonrigid materials since milling objectives are highly sensitive to changes in temperature and deformation. As a consequence, current robotic milling approaches designed for rigid materials may fail to ensure processing accuracy and security in such scenarios. To deal with this issue, this paper constructs a robotic milling simulation environment that incorporates a joint multi-fidelity surrogate model for nonrigid material milling with the consideration of temperature and deformation information. By means of reinforcement learning strategies to learn the knowledge of nonrigid material milling process, the elastic deformation can be effectively compensated and the temperature can be restricted within the preset bound while ensuring the processing efficiency. Finally, the effectiveness of the proposed control approach for nonrigid material milling is verified through simulation and experiment results.